CLEANING BIOFOULED MEMBRANES EARLY IMPROVES PLANT

OPERATION. COMPARISON OF AUTOPSY AND CLEANING RESULTS

ON A SMALL BWRO AND LARGE SWRO.

Stephen P. Chesters, Genesys International Ltd, 3A Aston Way Middlewich, UK CW10 0HS
Email: [email protected] Phone: 004 413 72 741 881
Matthew W. Armstrong, Genesys International Ltd, Middlewich, UK CW10 0HS
Maqsood Fazel, Genesys International Ltd, Middlewich, UK CW10 0HS

Abstract
It is a fact that all new reverse osmosis membranes start to foul as soon as they start operation.
The only variable is the rate at which they foul. Most RO plants operate at a differential pressure
(∆P) of 1.5 - 3 bar. Cleaning usually occurs when the ∆P is 10-15% above the design
specification. This paper explores the benefits of early identification of foulants through autopsy
allowing predictive maintenance cleaning of the plant to be conducted at the first signs of
fouling. This prevents a build up of difficult to remove deposits which reduce membrane
performance and life expectancy.

Autopsies were conducted on a 1,200 GPM (6,650m³/day) brackish water reverse osmosis plant
(BWRO) and a 22,000 GPM (120,000m³/day) sea water reverse osmosis plant (SWRO).
Foulants identified included biofilm, clay and iron. Different cleaning solutions were then tested
on samples of the actual membranes to determine the best cleaning protocol.

The authors present the results of autopsies, laboratory cleaning tests and full plant cleaning
results with a number of different cleaning agents. A strongly alkaline cleaner with a high ionic
strength appeared to be particularly effective at removing biofilm and clay. The authors theorise
that during the soaking process the high ionic strength of the cleaning solution on the feed side of
the membrane encourages natural osmosis of permeate water through the membrane to the feed
side thus causing some physical disruption to the biofilm and clay fouling layer. This allows the
detergent, surfactant and chelating molecules in the cleaner to break up the fouling layer
allowing a more effective clean. Once biofilm is removed then an acid clean proved much more
effective at removing iron deposits.

Comparisons are drawn between the results of cleaning early on the SWRO plant and late on the
BWRO plant. It is much easier to remove deposits before they have formed a substantial cake
layer resulting in more effective and less frequent cleaning which prolongs membrane life.

Introduction
Membrane manufacturers typically recommend that cleaning should be conducted when there is
a significant drop in performance. The standard advice has not been changed in over 30 years.
The DOW Filmtec Technical Manual states on Page 122 that: “Elements should be cleaned when
one or more of the below mentioned parameters are applicable:

1
• The normalized permeate flow drops 10%

• The normalized salt passage increases 5 - 10%

• The normalized pressure drop (feed pressure minus concentrate pressure) increases 10 -15%

If you wait too long, cleaning may not restore the membrane element performance successfully.
In addition, the time between cleanings becomes shorter as the membrane elements will foul or
scale more rapidly.” This is the same message offered in Hydranautics Technical bulletin of
1992: “If permeate flow drops, product water quality decreases, or applied pressure must be
increased to maintain normal productivity more than 10 to 15 percent, then the elements need to
be cleaned.” Analysis of the results of 500 membrane autopsies by Pena et al has shown that
60% of the main foulants are biological and particulate/colloidal fouling. [2] Clay and biofilm
tend to build up in layers and become compressed into the membrane surface if left to
accumulate. This paper explores the benefits of cleaning early when membrane foulants are
easier to remove.

Cleaning Early
It is a fact that all new reverse osmosis membranes start to foul as soon as they start operation.
The only variable is the rate at which they foul. Most RO plants operate at a differential pressure
(∆P) of 1.5 - 3 bar. Cleaning usually occurs when the ∆P is 10-15% above the design
specification. This is based on recommendations by membrane manufacturers. Figure 1 below
measures fouling with time. The perception threshold line indicates the time at which the
operator ‘perceives’ a change to system operation, whether this be a change in flux, pressure, or
salt passage. In reality as soon as water passes along the membrane (initial time = T0), the
membrane starts fouling. It is important to appreciate this fact; the membrane is fouling from
system start up, but nothing is observed until the fouling line crosses the perception line, at
which point the observer can see there is some change in the operation. The membrane acts like a
huge sponge with a myriad of active surfaces, so a significant amount of surface fouling can take
place before any operational change is observed.[1]

2
 Figure 1. Membrane fouling rate

Laboratory studies from numerous autopsies have confirmed that the sooner the membrane is
cleaned after fouling the easier it is to clean. If the foulant is not cleaned it becomes more
difficult to remove until finally the membrane cannot be cleaned at all. Clean 1 will theoretically
take place when perceived performance has dropped by 10-15%. The reality is that by the time
cleaning chemicals are sourced and a shutdown period is arranged the performance may have got
significantly worse. The clean may result in conditions returning to an acceptable level below the
fouling perception threshold. The clean after a delay will not however be as effective as it could
have been if performed earlier before more deposits added to the cake layer and became
compressed. Because the clean did not remove the majority of deposits the underlying fouling
rate is higher and the perception threshold of performance loss will be seen quicker and clean 2
conducted after a shorter period than may otherwise have been possible had the first clean
happened earlier.

The whole process can be planned by conducting autopsies on a membrane element from the
front and back of the plant. Foulants are most likely to appear at the front and scalants in the last
position element. Convincing some operators to remove and autopsy almost new elements may
be a struggle so smaller 2 or 4 inch sacrificial elements could easily be installed and used. The
autopsy will identify typical foulants. Once the foulants are known cleaning protocols can be
tested in the laboratory and appropriate cleaning agents can be stored on site ready for cleaning
as soon as there is a notable change in operating parameters. Any delay in cleaning will mean the
foulant will be become thicker and more compressed into the membrane surface and be much
more difficult to clean. Early maintenance cleaning of the plant prevents a build-up of difficult to
remove deposits which reduce membrane performance and life expectancy. Less time will be
required to conduct the cleaning and a more effective “deeper” clean can be achieved. This
means the subsequent fouling rate is lower, closer to the rate of the new membrane when

3
installed. The lower fouling rate reduces the frequency between subsequent cleans increasing
operational efficiency and further enhancing membrane lifespan.

Membrane Autopsy
In most cases, membranes are sent for autopsy because they have failed or are underperforming
significantly. That is salt passage, flow, flux and pressure differentials are outside of the design
specifications. The first stage of membrane autopsy is to record the element model, serial
number, position in the plant along with the chemical treatment regimes being applied. External
examinations and photographs are taken to record the physical condition. A vacuum test is
conducted to assess physical membrane damage. Samples can be taken from the feed inlet and
outlet. The outer casing is removed and membrane leaves unrolled in readiness for an internal
examination of the membrane sheets, glue lines, vexar and permeate carrier. Representative
deposit samples are taken from the membrane surface and photographs are taken. Membrane
samples are cut and prepared for characterisation and cleaning tests in a specially designed flat
sheet test rig. [Fig 2] Flow rate and salt rejection are determined and compared to design
specifications.

Figure 2. Flat sheet Test Rig

Physical damage to the membrane samples is conducted using the methylene blue dye test and
oxidation by halogens is tested using the Fujiwara method. The validity of the Fujiwara test has
been questioned in some quarters so additional tests using attenuated total reflectance infrared
spectroscopy (ATR/FTIR) are carried out. This can provide valuable information related to the
chemical structure of membrane and deposits. In the mid-infrared, absorption of radiation is
related to fundamental vibrations of the chemical bonds. Internal reflection spectrometry
provides information related to the presence of specific functional groups which can determine
the presence of halogens. Deposits are identified using elemental determination with the SEM-
EDXA system. This is based on analysis of X-rays produced via electron beam excitation of a
sample area. This technique allows analysis of a sample in selective areas. The limited depth of
analysis (typically a few microns), and the possibility to select the area of interest under the
electron beam, allows for local analysis to reveal differences in composition. The identification
and measurement of individual peak intensities in the X-ray spectrum can be done with a
computerized multichannel analyzer. In addition X- Ray diffraction is carried out to identify
crystalline solids by measuring the characteristic spaces between layers of atoms or molecules in

4
a crystal. Each crystalline solid has its unique characteristic X-ray powder pattern which may be
used as a "fingerprint" for its identification. X-ray photoelectron spectroscopy analysis (XPS) or
electron spectroscopy for chemical analysis (ESCA) is a quantitative spectroscopic technique
that measures the empirical formula, chemical state and electronic state of the elements that exist
within a material. XPS spectra are obtained by irradiating a material with a beam of X-rays while
simultaneously measuring the kinetic energy and number of electrons that escape from the top 1
to 10 nm of the material being analyzed. When the nature of the deposits on the membrane
surface is understood then several membrane samples are cleaned with a variety of cleaning
solutions at different pH and temperature to assess the best method for cleaning the plant. A full
cleaning protocol can then be recommended. The autopsy is written up as a complete report and
forwarded to the customer.

BWRO Plant Autopsy

The Ursus brewery in Buzau owned by SAB Miller is one of the biggest in Romania. The 1,200
GPM (6,500 m³/day) brackish water reverse osmosis plant is supplied via four boreholes and was
commissioned in 2002. The plant operated satisfactorily for three years and then increased iron
levels in the well water resulted in an increasing fouling rate. The RO plant membranes were
being cleaned every month with caustic and acid membrane cleaning chemicals. Results gave
only a slight decrease in differential pressure and increase in flow rate and permeate production.
The membranes had to be replaced on a two year cycle. In June 2008 the plant was surveyed and
a last element second stage membrane taken for autopsy. On the 26th June the autopsy report on a
Dow Filmtec BW 30LE-440 was issued. The membrane weighed 22.1 kg compared with a new
element weight of 15kg. The autopsy revealed a very heavy red-brown gelatinous deposit which
appeared on the feed inlet (Fig 3) and throughout the feed spacer and membrane surface.
Samples of deposit were taken from the membrane surface (Fig 4) and analysed using Scanning
Electron Microscopy – Energy Dispersive X-ray Analysis (SEM-EDXA) techniques to
determine elemental composition. The deposit consisted of iron combined with alumina-silicate
(clay) and a heavy biofilm. Bacteria and oxidation tests were also conducted revealing the
presence of aerobic bacteria. The membrane did not show signs of oxidation damage but flow
was significantly reduced. Cleaning tests were conducted on membrane samples and a powdered
high alkaline high ionic strength cleaner containing detergent, chelant and surfactants followed
by a mild acid cleaner gave the best results.

5
 Figure 3. Deposit on feed inlet

Figure 4. Deposit on membrane surface

Figure 5. SEM of bacteria and clay

BWRO Cleaning Results

In September 2008 the RO plant membranes were cleaned. A 2% solution of alkaline cleaner
(pH 11.8) was made up in the CIP tank and warmed up to 38°C and circulated through the
membranes. After 10 minutes the cleaning solution turned brown and cloudy and was discharged
and a fresh solution made up. Following circulation for 30 minutes the solution was left to soak
for 30 minutes and then flushed with permeate water. A large amount of deposits was removed
so the cleaning step was repeated a third time. Samples of each cleaning solution and rinse water
in Figure 6 demonstrate the progressive removal of biofilm and clay. A mild acid cleaner
containing surfactants was then circulated through the membranes. The cleaning solution turned
a red brown colour and fresh cleaning solution had to be made up three times circulated and
flushed out. Figure 7. The cleaning process was repeated three times over the next four months
until the membrane performance was restored. Subsequent cleaning cycles are now on a 6 month
basis and membranes have not been replaced for four years. Historically this plant suffered gross
biofilm, clay and iron fouling which had been allowed to build up over a long period; due to poor
understanding of the foulants and subsequent inefficient cleaning practice. Failure to fully

6
remove the deposits during CIP resulted in increased fouling potential causing significant
operating issues. Conventional cleaners were unsuccessful at removing deposits. The use of a
high ionic strength cleaner with extended periods of soaking managed to remove sufficient
biofilm and clay so that the acid stage of the cleaning could then take iron deposits into solution
and flush them out. This cleaning reaction may be due to permeate water flowing through the
membrane by direct osmosis during the soaking period. This would help dislodge deposits from
the feed side of the membrane allowing the cleaning chemicals to break up any layered deposits.
[3]

Figure 6. Cleaning solution from alkaline clean

Figure 7. Acid cleaning solution samples

SWRO Plant Autopsy

A very large 22,000GPM (120,000m³/day) sea water RO plant in the Middle East was
commissioned in October 2009. A programme of predictive maintenance was a key feature of
the operation of this plant. In March 2010 a Hydranautics SWC5 RO lead membrane element
from the first stage of the plant was sent for autopsy. There was no indication of any fouling
present from operational data but it was felt that an autopsy would identify typical foulants that
would ultimately accumulate on the membrane surfaces. A very thin gelatinous deposit was
found covering the membrane surface Figure 8. Analysis of the deposit revealed it consisted of
76.9% organic material in the form of protein derivatives relating to bacteria and biofilm
formation. An aerobic bacteria count showed a significant number of 7,300 colony forming units
per cm². Elemental analysis of the deposit also showed the presence of iron and aluminium-
silicate in the form of clay. The membrane samples were charachetrised using simulated sea
water with pH 7.2, Conductivity 49,000 µS/cm, and temperature of 25°C. Cleaning tests were
conducted on membrane samples measuring improvement in flow rate and salt passage of a
variety of cleaning chemistries. It was found that the use of a highly alkaline powder detergent
and chelant (cleaner A) showed superior cleaning efficacy when compared with a more

7
traditional programme of biocide (cleaner B) and traditional liquid alkaline cleaner (cleaner C).
The use of a warm caustic solution circulated for 2 hours showed a minor improvement in flow
rate of only 2.8%. The synergisitic effects of multiple chemistries and cleaning mechanisms in
one clening compound are well recognised. [5]

Figure 8: Gelatinous biofilm

deposit
Table 1: Summary of flat sheet test rig cleaning results

Sample Sodium Cleaner A Cleaner B & C

Hydroxide
Aerobic count 7,300 ND ND ND
CFU/cm²
Flow Rate 35.7 l/m²h +2.8% +39.7% +15.3%
Salt Rejection 99.6% 99.6% 99.7% 99.6%
pH NA 11.8 11.8 C – 11.8
Temp NA 38.7°C 38.6°C C- 38.7°C

SWRO Cleaning Results

After the first autopsy was conducted in March 2010 it was decided that cleaners A, B and C
would be available on site to conduct cleaning as soon as there was an early sign of fouling.
Differential pressure (DP) was monitored daily and a significant increase of 0.2 bar was
observed in rack 21 over a two week period in January 2011. The increase in DP was seen after a
leak on the shock chlorination system was fixed. While the pipework was leaking fouling rates
were low. It is possible that the resumption of shock chlorination actually created additional
nutrients for a proliferation in bacteria activity and subsequent biofilm growth leading to
increased fouling rate. Based on the autopsy and cleaning test results it was decided to initially
clean rack 21on 20th January 2011 with the traditional cleaning chemicals B & C namely a
biocide followed by an alkaline clean. As can be seen from the graph of DP in Figure 9 there was
an immediate reduction of 0.14 bar after the clean although the fouling rate remained the same.
By 1st March the DP had risen by 0.2 bar. The increase was deemed to be due to microbiological
fouling so a biocide clean was conducted. This did not achieve a reduction of DP but did seem to
halt fouling for 3 weeks at which point the underlying fouling rate resumed. Three weeks later
the DP had risen to 1.46 bar and at this point a biocide flush was followed by a full clean using

8
Cleaner A the powdered high alkaline, high ionic strength detergent, surfactant and chelant
cleaner. The cleaning agent was repeatedly circulated and soaked for 30 minute intervals and
flushing conducted with permeate. A much improved cleaning result was achieved with a
pressure drop of 0.3 bar as compared with 0.2 bar when using the conventional liquid cleaner.
Figure 10 shows pressure drops during a subsequent clean on 2nd July using just cleaner A which
shows enhanced results with reduction in DP of 0.54 bar. Significantly a deeper clean seems to
have resulted with DP reduced to 1 bar similar to that found during the first three months of plant
operation. It must be recognised that this plant was being cleaned as soon as a rise of 0.2 bar DP
was observed. Under normal circumstances operators would not start to react until a pressure
drop of over 1 bar. Brusilovsky et al comment on the Larnaca SWRO “where the delta P in an 8
element vessel after 4 years operation is 1.1 to 1.2 bar.” By reacting early with a predictive
cleaning protocol this large SWRO plant has been able to operate within specification despite a
challenging sea water source prone to red tide algal blooms.

Figure 9. Differential pressure following cleaning Nov 2010 – May 2011

9
 Figure 10. Differential pressure following cleaning May 2011 to Aug 2011

Conclusions

The recommendations by membrane manufacturers on when to clean and what chemistry to use
have changed very little in the last 30 years. The authors question the logic in waiting for
operating parameters to deteriorate by 10-15% before taking action. Results from a BWRO and
SWRO autopsy, cleaning tests and actual site cleans allow the following conclusions to be
drawn.

 If membranes are not cleaned early the deposit is very difficult to remove and may
require high ionic strength cleaners and periods of extensive soaking to remove biofilm
and clay.
 If predictive early cleaning is conducted then deposits are easier to remove and the
underlying fouling rate is lower.
 A membrane element autopsy or cartridge filter analysis within 3 months of start up will
identify typical and likely membrane foulants.
 Cleaning tests on the foulant enables a predictive cleaning protocol to be produced so the
necessary equipment and cleaning reagents can be sourced in advance speeding up the
reaction time to the first sign that fouling has taken place.

Predictive maintenance can now be adopted by the membrane cleaning industry to improve the
effective and efficient operation of sea and brackish water RO plant.

10
 References

[1] E. G. Darton - The role of chemicals in this thirsty planet IDAWC/PER11-262

World Congress/Perth Convention and Exhibition Centre September 4-9, 2011

[2] N. Peña, S. Gallego, F. del Vigo, S.P. Chesters. Evaluating impact of fouling on reverse
osmosis membrane performance. EDS Conf Barcelona 2012

[3] Liberman, B., Methods of direct osmosis membrane cleaning online for high SDI feed after
pretreatment, IDA Workshop,Tampa – San Diego, 22-26/03/04.

[4] Wui Seng Ang, Ngai Yin Yip, Alberto Tiraferri, Menachem Elimelech. Chemical cleaning of
RO membranes fouled by wastewater effluent: Achieving higher efficiency with dual-step
cleaning Journal of Membrane Science 382 (2011) 100– 106

[5] M. Brusilovsky, M. Faigon The impact of varying the number of elements per PV in SWRO
plants - actual and future Desalination 184(2005)Pages 233-240

11